U.S. government moves to restrict OpenAI’s GPT 5.6 release

The U.S. government is set to take control over which artificial intelligence models are released. According to a report by technology news publication The Information on Thursday, OpenAI’s upcoming GPT 5.6 model will be released only into a limited preview. Under this arrangement, the government will require approval “customer by customer” before a general release can occur. This move mirrors recent regulatory actions taken against rival AI lab Anthropic, whose models were pulled by regulators, signaling a broader shift toward direct federal oversight of the sector.

The potential delay introduces significant economic uncertainty for the industry. OpenAI Chief Executive Sam Altman has suggested that if the preview lasts a “couple of weeks,” it might not be a big problem. However, the precedent set by Anthropic suggests a much longer timeline. Two weeks after the U.S. government pulled Anthropic’s Fable and Mythos models, the industry remains in limbo. Anthropic’s Mythos model has been in preview for months with no indication of a general release. For AI labs spending billions to develop frontier models—highly capable AI models at the cutting edge of development—even a few weeks of regulatory delay could severely limit the economic upside of these costly systems, threatening their ability to improve their bottom lines.

This regulatory bottleneck has divided the tech industry, sparking intense finger-pointing between the major players. Industry conversations have devolved into bitter disputes, with some accusing Anthropic of running a regulatory capture scheme—which is a form of corruption where a regulatory agency acts in the commercial interest of the industry it is charged with regulating. Meanwhile, others are accusing OpenAI of cozying up to Trump to ice out a rival. Despite these accusations, the underlying issue remains a lack of clear regulatory standards. As GMU fellow and future OpenAI employee Dean Ball detailed in a post this morning, it remains entirely unclear what safety assurances or testing standards regulators are actually seeking to enforce, as the government currently lacks the expertise or capacity for this level of testing.

Why it matters

The U.S. government’s increasing control over AI model releases threatens to slow model development and impact the broader data center buildout, creating a bottleneck that affects all major AI labs.